https://thenewstack.io/nvidia-finally-adds-native-python-support-to-cuda/

TNS
OK
SUBSCRIBE
Join our community of software engineering leaders and aspirational
developers. Always stay in-the-know by getting the most important
news and exclusive content delivered fresh to your inbox to learn
more about at-scale software development.
EMAIL ADDRESS
REQUIRED
[                    ] SUBSCRIBE
RESUBSCRIPTION REQUIRED
 
It seems that you've previously unsubscribed from our newsletter in
the past. Click the button below to open the re-subscribe form in a
new tab. When you're done, simply close that tab and continue with
this form to complete your subscription.
RE-SUBSCRIBE
The New Stack does not sell your information or share it with
unaffiliated third parties. By continuing, you agree to our Terms of
Use and Privacy Policy.
Welcome and thank you for joining The New Stack community!
Please answer a few simple questions to help us deliver the news and
resources you are interested in.
FIRST NAME
REQUIRED
[                    ]
LAST NAME
REQUIRED
[                    ]
COMPANY NAME
REQUIRED
[                    ]
COUNTRY
REQUIRED
[Select ...                                  ]
ZIPCODE
REQUIRED
[                    ]
Great to meet you!
Tell us a bit about your job so we can cover the topics you find most
relevant.
What is your job level?
REQUIRED
[Select ...                              ] [                    ]
Which of these most closely describes your job role?
REQUIRED
[Select ...                           ] [                    ]
How many employees are in the organization you work with?
REQUIRED
[Select ...      ]
What option best describes the type of organization you work for?
REQUIRED
[Select ...                                                                                                 ]
[                    ]
Which of the following best describes your organization's primary
industry?
REQUIRED
[Select ...                                           ]
[                    ]
LINKEDIN PROFILE URL
 
[                    ]
Welcome!

We're so glad you're here. You can expect all the best TNS content to
arrive Monday through Friday to keep you on top of the news and at
the top of your game.

What's next?

Check your inbox for a confirmation email where you can adjust your
preferences and even join additional groups.

Follow TNS on your favorite social media networks.

        

Become a TNS follower on LinkedIn.

Check out the latest featured and trending stories while you wait for
your first TNS newsletter.

PREV
1 of 2
NEXT
VOXPOP
As a JavaScript developer, what non-React tools do you use most
often?

Angular
0%

Astro
0%

Svelte
0%

Vue.js
0%

Other
0%

I only use React
0%

I don't use JavaScript
0%
Thanks for your opinion! Subscribe below to get the final results,
published exclusively in our TNS Update newsletter:
[                    ]
SUBMIT
 
 
[                    ]
Search
More Results
ARCHITECTURE
Cloud Native Ecosystem Containers Databases Edge Computing 
Infrastructure as Code Linux Microservices Open Source Networking 
Storage
ENGINEERING
AI AI Engineering API Management Backend development Data Frontend
Development Large Language Models Security Software Development 
WebAssembly
OPERATIONS
AI Operations CI/CD Cloud Services DevOps Kubernetes Observability 
Operations Platform Engineering
PROGRAMMING
C++ Developer tools Go Java JavaScript Programming Languages Python 
Rust TypeScript
CHANNELS
Podcasts Ebooks Events Newsletter TNS RSS Feeds
THE NEW STACK
About / Contact Sponsors Advertise With Us Contributions
PODCASTS EBOOKS EVENTS NEWSLETTER CONTRIBUTE
ARCHITECTURE ENGINEERING OPERATIONS PROGRAMMING
Cloud Native Ecosystem Containers Databases Edge Computing 
Infrastructure as Code Linux Microservices Open Source Networking 
Storage
Perses Closes the Observability Gap with Declarative Dashboards
Mar 28th 2025 8:00am, by Nicolas Takashi
A Decade of Cloud Native: From CNCF, to the Pandemic, to AI
Mar 25th 2025 12:00pm, by Richard MacManus
Mastering AWS Lambda: Optimize Cost and Performance
Mar 19th 2025 2:00pm, by Ridge Kimani
Five Critical Shifts for Cloud Native at a Crossroads 
Mar 19th 2025 7:03am, by Kim McMahon
Download Our New Ebook on Cloud Sustainability
Mar 18th 2025 11:00am, by Vicki Walker
Leaner Development: How To Account for the Container Tax
Apr 1st 2025 12:00pm, by Steve Rodda
KubeCon EU 2025: Edera Protect Offers a Secure Container
Apr 1st 2025 8:00am, by Steven J. Vaughan-Nichols
NVIDIA Open Sources KAI Scheduler To Help AI Teams Optimize GPU
Utilization
Apr 1st 2025 1:00am, by Frederic Lardinois
KubeCon EU 2025: Mirantis' k0s and k0smotron Join CNCF Sandbox 
Mar 31st 2025 11:00am, by Steven J. Vaughan-Nichols
Endor: WebAssembly-Based Server in the Browser
Mar 29th 2025 1:00pm, by B. Cameron Gain
How To Generate 'AI' From a Database: Bruce Momjian
Mar 28th 2025 7:00am, by Joab Jackson
Valkey 8.1's Performance Gains Disrupt In-Memory Databases
Mar 25th 2025 6:06am, by Steven J. Vaughan-Nichols
Real-Time Write Heavy Workloads: Considerations and Tips
Mar 14th 2025 9:00am, by Felipe Cardeneti Mendes and Lubos Kosco
Can You Trust AI To Be Your Data Analyst? 
Mar 11th 2025 1:30pm, by Sarah Levy
Vector Search Without the Lock-In: Why Devs Like ClickHouse
Mar 11th 2025 8:30am, by Lewis DiFelice
Should You Care About Fermyon Wasm Functions on Akamai?
Mar 28th 2025 1:00pm, by B. Cameron Gain
Are Edge Computing and Cloud Computing in Competition?
Mar 27th 2025 6:00am, by Meredith Shubel
Edge Data Centers Offer Benefits for Remote Industrial Apps
Mar 26th 2025 8:00am, by Meredith Shubel
AI Is Coming to the Edge, but It Will Look Different
Mar 20th 2025 1:00pm, by Damir Mujezinovic
Unified Telemetry + Observability: The Future of Data Management
Mar 7th 2025 8:00am, by Lori Bertelli
IT Leaders Brace for Tariff Fallout on Infrastructure and Cloud Costs
Mar 26th 2025 10:00am, by Krishna Subramanian
How To Select the Best Cloud Model for Your AI Strategy
Mar 25th 2025 1:00pm, by Abu Hassan Peer
Why FinOps Belongs in Your CI/CD Workflow
Mar 25th 2025 5:00am, by Amit Liberman
NixOps Lives On: Introducing NixOps 4
Mar 21st 2025 9:00am, by B. Cameron Gain
AI Agents Are Dumb Robots, Calling LLMs
Mar 20th 2025 2:00pm, by Alex Williams
How Linux Kernel Deals With Tracking CVE Security Issues
Apr 1st 2025 7:00am, by Steven J. Vaughan-Nichols
Check Out Debian, the 'Mother of All Linux Distributions'
Mar 29th 2025 7:00am, by Jack Wallen
Akamai Picks Up Hosting for Kernel.org
Mar 27th 2025 10:00am, by Frederic Lardinois
Linux Kernel 6.14: Enhanced Drivers, Security, Performance
Improvements
Mar 25th 2025 8:15am, by Damon M. Garn
Cubic: Build a Custom Linux Distribution Based on Ubuntu
Mar 23rd 2025 7:00am, by Jack Wallen
Sandbox Testing: The DevEx Game-Changer for Microservices
Mar 27th 2025 7:11am, by Arjun Iyer
Testing Microservices: Message Isolation for Kafka, SQS, More
Mar 20th 2025 7:09am, by Arjun Iyer
The Million-Dollar Problem of Slow Microservices Testing
Mar 6th 2025 8:05am, by Arjun Iyer
Introduction to Service Mesh
Feb 22nd 2025 8:00am, by TNS Staff
Deno Creates Board, Charter for JavaScript Registry Project
Feb 15th 2025 5:00am, by Loraine Lawson
How Linux Kernel Deals With Tracking CVE Security Issues
Apr 1st 2025 7:00am, by Steven J. Vaughan-Nichols
Valkey 8.1's Performance Gains Disrupt In-Memory Databases
Mar 25th 2025 6:06am, by Steven J. Vaughan-Nichols
The New Look and Feel of Apache Kafka 4.0
Mar 24th 2025 7:28am, by Jelani Harper
Eclipse Theia: The 'DeepSeek' of AI Tooling?
Mar 20th 2025 10:00am, by Darryl K. Taft
Open Infrastructure Foundation Joins Forces With Linux Foundation
Mar 20th 2025 8:00am, by Steven J. Vaughan-Nichols
KubeCon EU 2025: Aviatrix's Enterprise Firewall for Kubernetes
Mar 24th 2025 5:00am, by Joab Jackson
NVIDIA Unveils Next-Gen Rubin and Feynman Architectures, Pushing AI
Power Limits
Mar 19th 2025 3:00pm, by Adrian Cockcroft
2.8 Million Reasons Why You Can't Trust Your VPN
Mar 17th 2025 9:00am, by Nick Taylor
AI in Network Observability: The Dawn of Network Intelligence
Mar 16th 2025 8:00am, by Christoph Pfister
Create a Samba Share and Use From in a Docker Container
Mar 9th 2025 6:00am, by Jack Wallen
Changed Block Tracking Is Here for Kubernetes Resilience
Mar 24th 2025 8:30am, by Mark Lavi
Real-Time Write Heavy Workloads: Considerations and Tips
Mar 14th 2025 9:00am, by Felipe Cardeneti Mendes and Lubos Kosco
Create a Samba Share and Use From in a Docker Container
Mar 9th 2025 6:00am, by Jack Wallen
Database Pricing Trends: What You Need to Know
Mar 4th 2025 11:00am, by Antoni Olendzki
Stop Freezing Your Data to Death
Feb 25th 2025 7:08am, by Franz Knupfer
AI AI Engineering API Management Backend development Data Frontend
Development Large Language Models Security Software Development 
WebAssembly
NVIDIA Finally Adds Native Python Support to CUDA
Apr 2nd 2025 2:00pm, by Agam Shah
NVIDIA Making Radical Changes to CUDA After Nearly 20 Years
Apr 1st 2025 2:00pm, by Agam Shah
Red Hat Arms Developers With AI-Powered Migration Tools
Apr 1st 2025 1:00pm, by Steven J. Vaughan-Nichols
Agentic AI and Platform Engineering: How They Can Combine 
Apr 1st 2025 11:00am, by Jennifer Riggins
AI Agents Must Learn From ChatGPT's Data Wrongs
Apr 1st 2025 10:00am, by Yukai Tu
Study: AI Turns Evil After Training on Insecure Code
Mar 27th 2025 2:00pm, by Kimberley Mok
Akamai Picks Up Hosting for Kernel.org
Mar 27th 2025 10:00am, by Frederic Lardinois
5 Non-AI Reasons Why You Still Have Toil
Mar 26th 2025 11:00am, by Leo Vasiliou
Developer's Guide to the Built-In Tools of OpenAI Agents SDK
Mar 26th 2025 9:12am, by David Eastman
How to Make Sure Process Debt Doesn't Kill Your AI Projects
Mar 25th 2025 9:00am, by Scott Willson
MCP: The Missing Link Between AI Agents and APIs
Mar 13th 2025 7:30am, by Richard MacManus
Apollo: GraphQL Now Connects to REST APIs With Little Fuss
Mar 11th 2025 7:30am, by B. Cameron Gain
MCP Bridges LLMs to the Apps They Need
Mar 7th 2025 11:30am, by Joab Jackson
The Deployment Bottleneck No One Talks About
Mar 3rd 2025 1:06pm, by Rak Siva
The Rise of AI Agents: How Arazzo Is Defining the Future of API
Workflows
Feb 28th 2025 9:00am, by Frank Kilcommins
JavaScript's Missing Link: Wasp Offers Full Stack Solution
Mar 26th 2025 2:00pm, by Loraine Lawson
Pagoda: A Web Development Starter Kit for Go Programmers
Mar 19th 2025 6:10am, by Loraine Lawson
Apollo: GraphQL Now Connects to REST APIs With Little Fuss
Mar 11th 2025 7:30am, by B. Cameron Gain
Introduction to Backend Development
Mar 10th 2025 12:30pm, by TNS Staff
Laravel Adds Inertia To Stack and Offers React Starter Kit
Mar 4th 2025 10:00am, by Loraine Lawson
AI Agents Must Learn From ChatGPT's Data Wrongs
Apr 1st 2025 10:00am, by Yukai Tu
Why Centralized AI Fails in Enterprise: The Case for a Federated
Architecture
Mar 28th 2025 10:00am, by Anant Bhardwaj
How To Generate 'AI' From a Database: Bruce Momjian
Mar 28th 2025 7:00am, by Joab Jackson
Tobiko Launches Its SQLMesh-Based Cloud Service Into GA
Mar 25th 2025 11:00am, by Frederic Lardinois
Observability Without a Data Lake Might No Longer Work
Mar 24th 2025 6:00am, by B. Cameron Gain
How AI Agents Are Quietly Transforming Frontend Development
Mar 31st 2025 2:00pm, by Alexander T. Williams
Researchers Find Next.js Middleware Vulnerability
Mar 29th 2025 3:01am, by Loraine Lawson
JavaScript's Missing Link: Wasp Offers Full Stack Solution
Mar 26th 2025 2:00pm, by Loraine Lawson
Brex's Web Dev Revamp Makes Code LLM Friendly, by Design
Mar 25th 2025 2:00pm, by Loraine Lawson
OpenAI API Now Supports Building Voice Agents
Mar 22nd 2025 5:00am, by Loraine Lawson
How Volcano Addresses LLM Training and Inference Challenges
Apr 1st 2025 9:30am, by Xuzheng Chang
ML and LLM Adoption Challenged Most Often by Observability
Mar 31st 2025 7:00am, by Lawrence E Hecht
Brex's Web Dev Revamp Makes Code LLM Friendly, by Design
Mar 25th 2025 2:00pm, by Loraine Lawson
How to Put Guardrails Around Containerized LLMs on Kubernetes
Mar 25th 2025 7:00am, by Sanjay Basu and Victor Agreda
AI Agents Are a Security Ticking Time Bomb
Mar 24th 2025 9:00am, by Alexander Borodetskiy
Shadow Testing Superpowers: Four Ways To Bulletproof APIs
Apr 2nd 2025 1:00pm, by Arjun Iyer
Bot Protection Strategies: Choosing the Right Approach for Your Stack
Apr 2nd 2025 11:00am, by Ron Northcutt
What AI Agents Do in the Shadows
Apr 2nd 2025 10:00am, by Ben Kliger
5 Kubernetes Vulnerabilities To Watch Out For
Apr 2nd 2025 7:00am, by Alex Handy
Google Extends Gmail Client-Side Encryption to All Users
Apr 1st 2025 9:00am, by Joab Jackson
Level Up Your Python: Higher-Order Functions Explained
Apr 2nd 2025 12:00pm, by Jessica Wachtel
SaaS Is Broken: Why Bring Your Own Cloud (BYOC) Is the Future
Mar 30th 2025 10:00am, by Noam Levy
Endor: WebAssembly-Based Server in the Browser
Mar 29th 2025 1:00pm, by B. Cameron Gain
Cascii and Why Developers Should Use ASCII Diagrams
Mar 29th 2025 11:00am, by David Eastman
Sprint Retrospectives: The Unsung Hero of Agile Success
Mar 29th 2025 10:00am, by Ben Grimwade
Endor: WebAssembly-Based Server in the Browser
Mar 29th 2025 1:00pm, by B. Cameron Gain
Should You Care About Fermyon Wasm Functions on Akamai?
Mar 28th 2025 1:00pm, by B. Cameron Gain
Akamai Picks Up Hosting for Kernel.org
Mar 27th 2025 10:00am, by Frederic Lardinois
Top 5 Uses of WebAssembly for Web Developers
Feb 16th 2025 7:00am, by Alexander T. Williams
The Feds Push for WebAssembly Security Over eBPF 
Feb 14th 2025 9:30am, by B. Cameron Gain
AI Operations CI/CD Cloud Services DevOps Kubernetes Observability 
Operations Platform Engineering
NVIDIA Finally Adds Native Python Support to CUDA
Apr 2nd 2025 2:00pm, by Agam Shah
What AI Agents Do in the Shadows
Apr 2nd 2025 10:00am, by Ben Kliger
NVIDIA Making Radical Changes to CUDA After Nearly 20 Years
Apr 1st 2025 2:00pm, by Agam Shah
AI Agents Must Learn From ChatGPT's Data Wrongs
Apr 1st 2025 10:00am, by Yukai Tu
NVIDIA Open Sources KAI Scheduler To Help AI Teams Optimize GPU
Utilization
Apr 1st 2025 1:00am, by Frederic Lardinois
Shadow Testing Superpowers: Four Ways To Bulletproof APIs
Apr 2nd 2025 1:00pm, by Arjun Iyer
Leaner Development: How To Account for the Container Tax
Apr 1st 2025 12:00pm, by Steve Rodda
Foundational Concepts in Platform Engineering 
Mar 27th 2025 1:00pm, by Manning Book Authors
Sandbox Testing: The DevEx Game-Changer for Microservices
Mar 27th 2025 7:11am, by Arjun Iyer
5 Non-AI Reasons Why You Still Have Toil
Mar 26th 2025 11:00am, by Leo Vasiliou
Google Extends Gmail Client-Side Encryption to All Users
Apr 1st 2025 9:00am, by Joab Jackson
SaaS Is Broken: Why Bring Your Own Cloud (BYOC) Is the Future
Mar 30th 2025 10:00am, by Noam Levy
How Disconnected Systems Drain SaaS Startups
Mar 27th 2025 11:00am, by Jeroen de Bruijn
Why Consolidating Observability Tools Is a Smart Move
Mar 26th 2025 3:00pm, by Gerardo Dada
IT Leaders Brace for Tariff Fallout on Infrastructure and Cloud Costs
Mar 26th 2025 10:00am, by Krishna Subramanian
Cascii and Why Developers Should Use ASCII Diagrams
Mar 29th 2025 11:00am, by David Eastman
Sprint Retrospectives: The Unsung Hero of Agile Success
Mar 29th 2025 10:00am, by Ben Grimwade
How Disconnected Systems Drain SaaS Startups
Mar 27th 2025 11:00am, by Jeroen de Bruijn
Why Consolidating Observability Tools Is a Smart Move
Mar 26th 2025 3:00pm, by Gerardo Dada
IT Leaders Brace for Tariff Fallout on Infrastructure and Cloud Costs
Mar 26th 2025 10:00am, by Krishna Subramanian
5 Kubernetes Vulnerabilities To Watch Out For
Apr 2nd 2025 7:00am, by Alex Handy
Has VMware Finally Caught Up With Kubernetes?
Apr 2nd 2025 6:03am, by B. Cameron Gain
KubeCon EU 2025: Kubernetes vNode, From the Makers of vCluster
Apr 2nd 2025 3:00am, by Joab Jackson
How Volcano Addresses LLM Training and Inference Challenges
Apr 1st 2025 9:30am, by Xuzheng Chang
How To Centralize Kubernetes Secrets Management With Vault
Apr 1st 2025 6:00am, by Utku Darilmaz
Observability in K8s: Moving From Reactive to Predictive
Apr 2nd 2025 9:00am, by Lori Bertelli
ML and LLM Adoption Challenged Most Often by Observability
Mar 31st 2025 7:00am, by Lawrence E Hecht
What eBPF Means for Observability vs. Security
Mar 31st 2025 6:17am, by B. Cameron Gain
SaaS Is Broken: Why Bring Your Own Cloud (BYOC) Is the Future
Mar 30th 2025 10:00am, by Noam Levy
Perses Closes the Observability Gap with Declarative Dashboards
Mar 28th 2025 8:00am, by Nicolas Takashi
Bot Protection Strategies: Choosing the Right Approach for Your Stack
Apr 2nd 2025 11:00am, by Ron Northcutt
Observability in K8s: Moving From Reactive to Predictive
Apr 2nd 2025 9:00am, by Lori Bertelli
How To Put a GUI on Ansible, Using Semaphore
Apr 1st 2025 6:00am, by Jack Wallen
KubeCon EU 2025: Komodor's Fix for Configuration Drift
Mar 31st 2025 5:00am, by Joab Jackson
Agentic AI: The Missing Piece in Platform Engineering
Mar 28th 2025 12:00pm, by Emilio Salvador
How To Hire a Platform Engineer
Apr 2nd 2025 8:00am, by Jennifer Riggins
Agentic AI and Platform Engineering: How They Can Combine 
Apr 1st 2025 11:00am, by Jennifer Riggins
Agentic AI: The Missing Piece in Platform Engineering
Mar 28th 2025 12:00pm, by Emilio Salvador
Foundational Concepts in Platform Engineering 
Mar 27th 2025 1:00pm, by Manning Book Authors
What's the Future of Platform Engineering?
Mar 27th 2025 9:00am, by Heather Joslyn
C++ Developer tools Go Java JavaScript Programming Languages Python 
Rust TypeScript
Memory-Safe C: TrapC's Pitch to the C ISO Working Group
Mar 12th 2025 7:00am, by David Cassel
Bjarne Stroustrup on How He Sees C++ Evolving
Mar 7th 2025 6:00am, by David Cassel
Curl's Daniel Stenberg on Securing 180,000 Lines of C Code
Feb 23rd 2025 6:00am, by David Cassel
Introduction to C++ Programming Language
Feb 11th 2025 6:00am, by TNS Staff
Nothing Janky About This New Programming Language
Jan 29th 2025 10:00am, by Darryl K. Taft
Red Hat Arms Developers With AI-Powered Migration Tools
Apr 1st 2025 1:00pm, by Steven J. Vaughan-Nichols
Developer's Guide to the Built-In Tools of OpenAI Agents SDK
Mar 26th 2025 9:12am, by David Eastman
Choose the Right Blazor Data Grid for Seamless Data Management
Mar 21st 2025 12:04pm, by Deyan Kamburov
Eclipse Theia: The 'DeepSeek' of AI Tooling?
Mar 20th 2025 10:00am, by Darryl K. Taft
Platform Engineers Must Have Strong Opinions
Mar 19th 2025 1:00pm, by Doug Sillars
Pagoda: A Web Development Starter Kit for Go Programmers
Mar 19th 2025 6:10am, by Loraine Lawson
Microsoft TypeScript Devs Explain Why They Chose Go Over Rust, C#
Mar 18th 2025 7:00am, by David Cassel
Go Power: Microsoft's Bold Bet on Faster TypeScript Tools
Mar 12th 2025 1:00pm, by Darryl K. Taft and Loraine Lawson
Introduction to Go Programming Language
Jan 17th 2025 9:00am, by TNS Staff
Golang Pub/Sub: Why It's Better When Combined With GoFr
Aug 28th 2024 6:00am, by Robert Kimani
Java Modernizes: New Tools for AI and Quantum Age
Mar 26th 2025 7:00am, by Chris J. Preimesberger
Oracle Ships Java 24: 'AI Is So Yesterday' Says VP
Mar 19th 2025 10:00am, by Darryl K. Taft
Vulnerability-Free Java Containers: A Practical Guide
Mar 8th 2025 9:00am, by Eric Murphy
AI Tools Now Essential in Java Dev's Productivity Arsenal
Mar 6th 2025 3:00pm, by Darryl K. Taft
2025 Is the Last Year of Python Dominance in AI: Java Comin'
Feb 7th 2025 11:00am, by Darryl K. Taft
Researchers Find Next.js Middleware Vulnerability
Mar 29th 2025 3:01am, by Loraine Lawson
JavaScript's Missing Link: Wasp Offers Full Stack Solution
Mar 26th 2025 2:00pm, by Loraine Lawson
Brex's Web Dev Revamp Makes Code LLM Friendly, by Design
Mar 25th 2025 2:00pm, by Loraine Lawson
OpenAI API Now Supports Building Voice Agents
Mar 22nd 2025 5:00am, by Loraine Lawson
Deno's Response to Node's Recent Support for TypeScript
Mar 15th 2025 6:00am, by Loraine Lawson
Level Up Your Python: Higher-Order Functions Explained
Apr 2nd 2025 12:00pm, by Jessica Wachtel
Rust Gets Its Missing Piece: Official Spec Finally Arrives
Mar 28th 2025 3:00pm, by Darryl K. Taft
Java Modernizes: New Tools for AI and Quantum Age
Mar 26th 2025 7:00am, by Chris J. Preimesberger
Master the Art of Python Debugging With These Tips
Mar 23rd 2025 8:00am, by Jessica Wachtel
Python's Built-In String Tools Every Developer Needs
Mar 20th 2025 11:00am, by Jessica Wachtel
NVIDIA Finally Adds Native Python Support to CUDA
Apr 2nd 2025 2:00pm, by Agam Shah
Level Up Your Python: Higher-Order Functions Explained
Apr 2nd 2025 12:00pm, by Jessica Wachtel
Insert Data into a MySQL Database via a Python Script
Mar 26th 2025 5:21am, by Jack Wallen
Master the Art of Python Debugging With These Tips
Mar 23rd 2025 8:00am, by Jessica Wachtel
Top 10 NLP Tools in Python for Text Analysis Applications
Mar 21st 2025 10:52am, by Alexander T. Williams
Microsoft TypeScript Devs Explain Why They Chose Go Over Rust, C#
Mar 18th 2025 7:00am, by David Cassel
Go Power: Microsoft's Bold Bet on Faster TypeScript Tools
Mar 12th 2025 1:00pm, by Darryl K. Taft and Loraine Lawson
Oracle Won't Release 'JavaScript' Without a Fight
Jan 11th 2025 5:00am, by Loraine Lawson
The Year in JavaScript: Top JS News Stories of 2024
Dec 27th 2024 6:30am, by Loraine Lawson
How OOP Developers Can Get To Know TypeScript Through Deno
Oct 19th 2024 8:00am, by David Eastman
2025-04-02 14:00:14
NVIDIA Finally Adds Native Python Support to CUDA
AI / AI Operations / Python

NVIDIA Finally Adds Native Python Support to CUDA

For years, NVIDIA's CUDA software toolkit for GPUs didn't have native
Python support. But that's now changed.
Apr 2nd, 2025 2:00pm by Agam Shah
Featued image for: NVIDIA Finally Adds Native Python Support to CUDA

In 2024, Python became the most popular programming language in the
world -- overtaking JavaScript -- according to GitHub's 2024 open
source survey.

For years, NVIDIA's CUDA software toolkit didn't have native Python
support. But that's now changed.

At GTC, NVIDIA announced native support and full integration of
Python in its CUDA toolkit. Developers will be able to use Python for
direct execution of algorithmic-style computing on GPUs.

"We've been working hard to bring accelerated Python, first class,
into the CUDA stack," said Stephen Jones, CUDA architect, during a
presentation at the recent GTC conference.

For programmers, the implications are massive. CUDA was born from C
and C++, and now coders don't need knowledge of those programming
languages to use the toolkit.

"Python for CUDA should not look like C. It should look like Python,"
Jones said.

    "Python CUDA is not just C translated into Python syntax."
    -- Stephen Jones, CUDA architect

Coders can use natural Python interfaces and the scripting model of
calling functions and libraries to create AI programs for execution
on NVIDIA GPUs.

"Python CUDA is not just C translated into Python syntax. It's got to
be something which is natural to a Python developer," Jones said.

Native Python Opens New Doors

NVIDIA's native support for Python on CUDA opens the developer
toolkit to millions of developers. CUDA previously required
developers to know C++ or Fortran. The programming kit had some
Python tooling, but it wasn't natively supported.

The number of CUDA users was only 4 million in 2023, up from 2
million in 2020, according to The Futurum Group. But Python is the
fastest-growing language in the world. NVIDIA will get access to
millions of Python coders -- especially in developing countries such
as India and Brazil, where coders make fervent contributions to open
source projects.

Python support will also make NVIDIA's infrastructure ready to use in
emerging markets. A bulk of NVIDIA GPUs are in the U.S. and Europe,
but telecom and infrastructure companies in India are building out
major GPU installations that will be operational in the coming years.

NVIDIA is pouring fuel into recruiting programmers and wants to
support more programming languages, including Rust and Julia.

How Pythonic CUDA Was Built

CUDA includes libraries, SDKs, compilers, host runtimes, tools, and
pre-packaged software and algorithms. NVIDIA has added pieces to the
entire Pythonic CUDA stack.

NVIDIA's focus was to provide GPU acceleration without getting out of
Python. CUDA Python can't be just kernel offerings; it needs
everything across the stack and a smooth execution flow, Jones said.

"You have to be able to write a kernel and drop it into PyTorch, but
you also have to be able to call Pythonic libraries and all of these
other things," Jones said.

Effectively, there's nothing at the compiler layer, as it's built
around just-in-time (JIT) compilation. That significantly reduces the
number of dependencies in the tree for GPUs in the stack.

"Keeping this interoperability between all the layers is going to be
a huge gain for productivity and being able to use Python
end-to-end," Jones said.

Initially, NVIDIA built base Python bindings (which include the
runtime compiler) and Python libraries such as cuPyNumeric, which is
a drop-in replacement for NumPy, the most widely used computational
library in Python. cuPyNumeric changes just one import directive, and
NumPy code goes from running on the CPU to running on a GPU.

    CUDA Core is a "Pythonic reimagining of the CUDA runtime to be
    naturally and natively Python."
    -- Stephen Jones, CUDA architect

Over the last year, NVIDIA made CUDA Core, which Jones said is a
"Pythonic reimagining of the CUDA runtime to be naturally and
natively Python."

CUDA Core has the execution flow of Python, which is fully in process
and leans heavily into JIT compilation.

"You should not be dropping out command-line compilers or anything
like that, you should be fully in process," Jones said, adding that
this significantly reduces the number of dependencies in the tree for
GPUs in the stack.

NVIDIA created a library called NVMath Python with unified interfaces
for both host-side and device-side library calls. The ability to fuse
library calls brings big performance improvements, Jones said.

The company has also built libraries that can access accelerated C++
libraries directly from Python code.

"Because this sits on top of the underlying infrastructure that we've
built over the years... we didn't re-implement these in Python. We made
sure that it links in the underlying fine-tuned C++ code so [that]
your performance difference is negligible," Jones said.

NVIDIA also added tools for profilers and code analyzers.

The Programming Model

Python makes coding straightforward and coders don't have to worry
much about the underlying hardware. With that in mind, NVIDIA is
adding a coding layer that aligns with higher-level abstraction for
execution on GPUs.

The new programming model, called CuTile interface, is being
developed first for Pythonic CUDA with an extension for C++ CUDA
coming later.

CuTile is "fundamentally more platonic," as Python programmers today
think more in terms of arrays than threads (which is more of a C++
trait).

Developers can't magically take Python code and export it for GPU
acceleration. CUDA typically takes a problem and breaks it into
thousands of smaller blocks that are processed separately on GPUs.

The blocks are broken up into smaller tiles, which run thousands of
threads processing single elements. The threads gang up into a single
operation.

The ability to go all the way down to process single elements at the
thread level in parallel gives the GPUs their massive computing
power.

But NVIDIA figured that GPU execution doesn't need to go all the way
down to the thread level. Processing can also be done midway at the
level of tiles, which is where the CuTile programming model fits in.

    Unlike C++, Python isn't granular by design.

CuTile does an efficient job of mapping arrays to GPUs at a less
granular level, which makes the code easier to understand and debug.
"And fundamentally, it comes out to the same performance," Jones
said.

Data in tiles could be structured as vectors, tensors, or arrays.
Compilers can do a better job of thread mapping whole array
operations from one block of threads to GPUs.

"Very often the compiler will do better than I can do because the
compiler deeply understands what I'm doing... [and] the fine details of
how the GPU runs," Jones said.

Unlike C++, Python isn't granular by design.

"There's a lot of these things out there. OpenAI's Triton would be a
good example. And I think that those are a natural fit for Python
programs," Jones said.

TRENDING STORIES
 
YOUTUBE.COM/THENEWSTACK
Tech moves fast, don't miss an episode. Subscribe to our YouTube
channel to stream all our podcasts, interviews, demos, and more.
SUBSCRIBE
Created with Sketch.
 
[501d027b-a]
Agam Shah has covered enterprise IT for more than a decade. Outside
of machine learning, hardware and chips, he's also interested in
martial arts and Russia.
Read more from Agam Shah
SHARE THIS STORY
     
TRENDING STORIES
SHARE THIS STORY
     
TRENDING STORIES
TNS DAILY NEWSLETTER Receive a free roundup of the most recent TNS
articles in your inbox each day.
[                    ] SUBSCRIBE
The New Stack does not sell your information or share it with
unaffiliated third parties. By continuing, you agree to our Terms of
Use and Privacy Policy.
ARCHITECTURE
Cloud Native Ecosystem Containers Databases Edge Computing 
Infrastructure as Code Linux Microservices Open Source Networking 
Storage
ENGINEERING
AI AI Engineering API Management Backend development Data Frontend
Development Large Language Models Security Software Development 
WebAssembly
OPERATIONS
AI Operations CI/CD Cloud Services DevOps Kubernetes Observability 
Operations Platform Engineering
CHANNELS
Podcasts Ebooks Events Newsletter TNS RSS Feeds
THE NEW STACK
About / Contact Sponsors Advertise With Us Contributions
 
roadmap.sh

Community created roadmaps, articles, resources and journeys for
developers to help you choose your path and grow in your career.

Frontend Developer Roadmap Backend Developer Roadmap Devops Roadmap
(c) The New Stack 2025
Disclosures Terms of Use Advertising Terms & Conditions Privacy
Policy Cookie Policy
FOLLOW TNS
        
* *